1. Field of the Invention
The present invention relates generally to a location measurement method and, in particular, to a location measurement method of a mobile node using a predictive filter that is capable of improving the location measurement accuracy of the mobile node.
2. Description of the Related Art
In a method for calculating location measurement error of a mobile node, a predictive filter is used to predict its location using the past measurement information and environmental information.
In order to compensate location measurement error of a mobile node, a predictive filter based compensation method using past measurements and environmental information, or a finger printing method using previously recorded site-specific signal patterns are used.
The predictive filter allows estimation of a current location of the mobile node using the past location and state information and compensates for the offset between the measured and predicted locations by adopting weights depending on the reliability of the information. A Kalman filter is one of the well known predictive filters. The Kalman filter adjusts a weight applied to a estimated value of a covariance matrix and an actually measured value.
The Kalman filter-based location prediction method is briefly explained as follows. First, a mobile node acquires coordinates of at least three anchor nodes and measures periodically the distances from the anchor nodes and then determines coordinates that have the smallest errors from each anchor node through a least square method as the location of the mobile node. The acquired coordinates are input to the Kalman filter so as to be output as a compensated coordinates. This method is described in more detail as follows, with reference to FIGS. 1, 2A, and 2B.
FIG. 1 is a graph illustrating a simulated result of measurement error compensation using a conventional least square method and Kalman filter, and FIGS. 2a and 2b are conceptual diagrams illustrating exemples of movements of a mobile node in indoor environments.
In FIG. 1, the axes x and y denote distances in units of meters (m). The measurement period (Δt) is set to 2 seconds, and the mobile node moves at a speed of 0.6 m/s.
The solid line 110 denotes the movement line of the mobile node, and mark x 130 denotes a location value estimated by using the least square method, the bold solid line 150 denotes a movement line of the mobile node compensated for the location measurement errors using the Kalman filter.
Here, it is assumed that the weight of the Kalman filter, i.e. process covariance matrix, and the measurement covariance matrix are constant.
RMSE (LeastSquare) is a Root Mean Squared Error (RMSE) obtained by applying the least square method. That is, the RMSE indicates the amount of mean error between the coordinates calculated by the least square method and the measured coordinates. RMSE (KalmanFilter) is the RMSE obtained by applying the Kalman filter. That is, the RMSE (KalmanFilter) indicates the size of a mean error between the coordinates obtained by applying the least square method and then the Kalman filter and the measured coordinates.
As shown in FIG. 1, when a constant covariance is used and the movement direction of the mobile node changes abruptly, the route obtained by applying the Kalman filter deviates from the actual movement route. This effect can occur when a movement direction of a mobile node previously moving straight along a corridor changes its movement direction at a corner of the corridor (see FIG. 2a), or when the mobile node enters a room positioned at one side of the corridor (see FIG. 2b).
In order to solve this problem, research has been focused on searching for ways to determine the weight of the predictive filter, adaptive to the situation of the mobile node. As a result, the most of the conventional techniques have been developed with a supplementary device such as an acceleration sensor and/or an initia sensor for acquiring information required for adjusting the weight of the predictive filter.
However, the conventional techniques are disadvantageous since additional physical elements, such as the acceleration sensor and initia sensor, increase the manufacturing cost and size of the mobile node. Furthermore, processing the information acquired by the additional elements increases the system complexity.